LangChain MCP Adapters Skill

Expert assistance for langchain-mcp-adapters: the official LangChain bridge to MCP (Model Context Protocol) servers. Converts MCP tools, prompts, and resources into LangChain-native objects usable by any LangChain agent or chain.

Install: pip install langchain-mcp-adapters

Reference: references/api.md (500 KB — full API reference) and references/llms.md (28 KB — index).

When to Use This Skill

Activate when:

• Connecting to MCP servers — using MultiServerMCPClient or create_session() to connect via stdio, SSE, HTTP, or WebSocket
• Loading MCP tools — calling load_mcp_tools() or client.get_tools() to get BaseTool-compatible tools
• Configuring connection types — choosing between StdioConnection, SSEConnection, StreamableHttpConnection, WebsocketConnection
• Adding tool interceptors — implementing ToolCallInterceptor for retry, caching, rate limiting, or auth
• Handling MCP callbacks — using LoggingMessageCallback, ProgressCallback, or ElicitationCallback
• Loading MCP resources or prompts — calling load_mcp_resources(), get_mcp_resource(), or load_mcp_prompt()
• Prefixing tool names — avoiding name collisions across multiple MCP servers with tool_name_prefix=True
• Converting to FastMCP — using to_fastmcp() to expose LangChain tools as a FastMCP server